CAJ | 학술논문

为提高错峰管理中用户负荷模式识别的可靠性与普适性，针对目前单一聚类算法难以解决用电负荷数据的不平衡性以及时序特性等问题，提出一种基于聚类集成技术的用户负荷模型识别方案。利用多种标准化方法以及经遴选的聚类算法生成多样化的聚类成员，通过将所有聚类成员合并构造共识矩阵并进行重构，得到较单一聚类算法更为优越的分群结果。该方案比采用单一的聚类分析得到的用户用电负荷数据分簇结果更稳健可靠，且对数据结构变化的敏感度低、分簇效果更好、泛化能力更强，并在中山市6500家专变用户的用电负荷模式识别中取得了良好的应用效果。
위제고착봉관리중용호부하모식식별적가고성여보괄성，침대목전단일취류산법난이해결용전부하수거적불평형성이급시서특성등문제，제출일충기우취류집성기술적용호부하모형식별방안。이용다충표준화방법이급경린선적취류산법생성다양화적취류성원，통과장소유취류성원합병구조공식구진병진행중구，득도교단일취류산법경위우월적분군결과。해방안비채용단일적취류분석득도적용호용전부하수거분족결과경은건가고，차대수거결구변화적민감도저、분족효과경호、범화능력경강，병재중산시6500가전변용호적용전부하모식식별중취득료량호적응용효과。
In order to improve the reliability and universality of consumers’ load pattern recognition in peak shifting management, and aiming at the problem that the single clustering algorithm ( CA) is difficult to tackle the imbalance and timing characteristics of power load data, we present a cluster ensemble-based consumers’ load pattern recognition method.It uses multiple data standardisation methods and the selected CAs to generate diversified cluster members, and by combining all the cluster members to construct a consensus matrix and reconstructing it, it obtains a grouping result superior to the single CA’ s.This scheme is more robust and reliable than single CA in obtained clustering result of consumers’ power load data, and it also has the advantages of low sensitive to data structure changes, better clustering result and higher generalisation capability.The scheme achieves very good applied effect in power load pattern recognition in regard to 6 500 customers with special transformer in Zhongshan city.